Off the shelf
Target prediction, molecular docking and ADMET prediction are public and freely available. We use them.
EP-001: mechanism, evidence and next steps
In independent CRO studies, EP-001’s test articles improved five endpoints in human neuronal cells under high-glucose stress. Computational screening maps the lead molecules to eight targets in the neuropathy cascade: six with published evidence in diabetic neuropathy, two in glycaemic control upstream of it.
Where existing drugs stop
Diabetic neuropathy sits downstream of several interacting pathways: oxidative stress, mitochondrial failure, inflammation and glycation. First-line drugs act on the pain signal instead.
Our lead molecules were first identified in a single botanical source. From them we are designing new chemical entities: synthetic analogues, patentable as new molecules, for an oral drug that protects the nerve.
Earthpharm NPDE
How we know which molecules matter.
A single botanical source contains hundreds of secondary metabolites. Establishing that it works is one problem. Knowing which molecules do the work, and which we haven’t found yet, is a different one, and it is the problem Earthpharm NPDE, our Natural Product Discovery Engine, is being built for.
Our workflow runs high-resolution mass spectrometry against target prediction, pathway mapping and developability modelling, then ranks candidates against the biology of this specific disease. Every ranking is re-weighted by what the bench returns.
Trained on the wrong chemistry
Those models learn from synthetic, drug-like molecules. Natural products are more complex, more polar, and absorbed by routes the models do not represent, so they flag our leads as undevelopable. The published pharmacokinetics say otherwise.
What we are building
The measured natural-product dataset that corrects them, generated in our own laboratory, and the scoring layer trained on it.
A prediction is a hypothesis, not evidence of binding or efficacy. Every ranked candidate goes to the bench, and we report which predictions held.
Multi-target computational panel run across eight upstream targets. Binding profiles, scoring methodology and per-target mechanistic detail are available under NDA.
How the engine worksComputational target panel
Computational screening maps EP-001’s lead molecules to eight upstream targets. Two, α-glucosidase and α-amylase, have bench enzyme-inhibition data. The other six are predictions; confirming them at the bench is part of the programme.
Direct diabetic neuropathy literature
Glycaemic control, upstream of neuropathic progression
Citations link to open-domain literature establishing each target’s role. Earthpharm’s binding profile and per-target mechanistic detail are reserved for partner conversations under NDA.
Independent CRO data in human neuronal cells
Five purified lead molecules, each tested alone, plus a reference compound. Human neuronal cells (SH-SY5Y) under 30 mM glucose for 48 hours, treated at 100 µg/mL. Ranges across the five molecules against the high-glucose control; apoptosis and oxidative stress against the lowest stressed-control value.
1.8 to 2.4 ×
Neuronal viability, increased
10 × or more
Apoptosis, reduced
2.2 to 2.4 ×
Oxidative stress, reduced
2.1 to 3.1 ×
Neurite outgrowth, increased
2.5 to 4.8 ×
Mitochondrial membrane potential, increased
Not yet shown: any animal or human data. A blinded data dossier is available on request; the full dataset under NDA.
Oral formulation
The oral formulation is developed for the lead analogue once it is chosen, alongside the animal studies.
Where the programme is now
Preclinical, TRL 3 to 4.
- Months 1 to 9
- Design and make the analogues, test them in cells, and file provisional patents.
- Months 9 to 24
- Animal studies, a 3D organoid model and the oral formulation, and a data package for pharma partners.
- After that
- GLP toxicology and the IND-enabling package, then a clinical trial application with a partner.
Identities are coded pending patent filings. Available under NDA.